Abhinav Nandwani

Applied AI Labs @ AMD · B.S. Computer Engineering & Computer Science, UW–Madison

I work in Applied AI Labs at AMD, where we bring AI into how chips get designed. A lot of my time goes into seeing how well coding assistants hold up on real engineering work—and making sure we can check their work with simulations and tests so the feedback actually means something. I’ve also shipped debugging tools that help designers read simulation waveforms faster, and I contribute to evaluating external AI vendors and startups building tools for chip design.

Before AMD I researched in Professor Daifeng Wang’s lab at UW–Madison, working on machine learning for genomics at large scale: running models on big GPU clusters and adapting them efficiently to new datasets.

These days I’m drawn to ML systems: training and serving models reliably at scale, the tooling that makes agents usable in real workflows, and the boundary where hardware, compilers, and model stacks have to line up for performance to land.

Abhinav Nandwani

Posters

Poster: fault-aware vision transformers for mission-critical applications, with Professor Parmesh Ramanathan

UW–Madison ECE is building a community of undergraduate researchers—symposia, fellowships, and community among labs. I presented at the first symposium in spring 2025.

With Professor Parmesh Ramanathan: Towards Fault-Aware AI Vision Transformers for Mission-Critical Applications—injecting GPU-style soft-error faults into a SegFormer semantic segmentation model on Cityscapes and measuring mIoU. Safety-relevant classes like car and person take some of the biggest hits when faults corrupt intermediate activations.

Recent work